| The accurate and effective emission data of pollutants is the primary prerequisite for pollution control.However,due to systematic and artificial fraud causes a large number of abnormal data during the actual monitoring process.This paper introduces the monitoring system of pollutant emission,analyzes the main sources of abnormal data and divides them into two major categories:system reason and artificial fraud.On the analysis of artificial counterfeit cases,two mathematical descriptions of abnormal data caused by artificial fraud are established:the limit model and the dilution model.Based on the ensemble of support vector machine method CeBag-SVM algorithm,similarity calculation method(MSET)is introduced to improve the combination method of base classifier from simple voting mechanism to weighted combination method,the new WCeBag-SVM algorithm.And the five kinds of data in the UCI dataset were used to test the performance of the algorithm.Through the research and analysis of the limestone-gypsum wet FGD process in coal-fired power plant,the model of desulfurization system is established.Finally,the data of sulfur dioxide emission from a thermal power plant for 90 consecutive days are validated.The experimental results show that this method can establish More accurate desulfurization system model.Finally,a semi-supervised learning method is proposed to detect abnormal data of pollutant emission.It is validated by using limits and dilution models to generate abnormal data on the validation dataset.The simulation results show that the proposed method based on semi-supervised learning has a good detection effect on abnormal data of pollution discharge caused by artificial counterfeiting and is a feasible method. |
